\documentclass[11pt]{article}
\usepackage{amsmath, amsthm, amssymb}
\usepackage{setspace}
\usepackage{graphicx}
\usepackage{multirow}
\usepackage{pdflscape}
\usepackage{harvard}
\usepackage{chngpage}
\usepackage{longtable}
\usepackage{url}
\usepackage{xcolor} 

\addtolength{\oddsidemargin}{-.5in}
\addtolength{\evensidemargin}{-.5in}
\addtolength{\textwidth}{1.0in}

\addtolength{\topmargin}{-.5in}
\addtolength{\textheight}{1.0in}

\newtheorem{claim}{Claim}
\newtheorem{prop}{Proposition}

\long\def\symbolfootnote[#1]#2{\begingroup%
\def\thefootnote{\fnsymbol{footnote}}\footnote[#1]{#2}\endgroup}

\usepackage[T1]{fontenc} 
\usepackage{lmodern} 

\begin{document}
	
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% TITLE PAGE %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
	
\begin{titlepage}
	
\begin{center}
\Large{Tables and Figures From: \\ \textbf{Preprimary Education and Early Childhood Development:  Evidence from Government Schools in Rural Kenya}}
\end{center}

\vspace{0.4in}	
	
\begin{center}
\begin{large}
			Original Authors: \\ Pamela Jakiela, Owen Ozier, 
			Lia C.~H.~Fernald, and Heather A.~Knauer
				
			
\end{large}	
\end{center}
	
	
\begin{center}
\begin{large}
			Replication Files Compiled \today
\end{large}
\end{center}
	
	
	\vspace{0.8in}
	
\end{titlepage}

\addtocounter{page}{1}


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% 

\doublespacing

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% 

\begin{figure}
\caption{Distance from the Primary School, Enrollment, and Luo Vocabulary} \label{fig:enroll-scatter}
	\begin{center}
		
		Panel A. School Enrollment 
		
		\includegraphics[width=0.64\textwidth]{out/F1A-lpoly-enroll.pdf}
		
		\vspace{0.2in}
		
		Panel B. Receptive Vocabulary in Luo
		
		\includegraphics[width=0.64\textwidth]{out/F1B-lpoly-luovocab.pdf}
	\end{center}
	
\noindent
\footnotesize{Data from 1,100 households with a child aged three or six years.  Figures plot local polynomial regressions of school enrollment (Panel A) or Luo receptive vocabulary (Panel B, measured in age-normalized z-scores) on distance to the local school (measured in kilometers).  Scatter plots depict averages in bins containing approximately fifty households. Shaded areas are 95\% confidence intervals. We do not cluster the confidence intervals by household in this figure, but only 12 households (out of 622 total) have more than one child in the sample.}  
\end{figure}

\clearpage



\begin{table}
\caption{Distance from School and Preprimary Enrollment} \label{tab:firststage}
\begin{center}
	\begin{small}
		\renewcommand{\arraystretch}{1.6}
		\begin{tabular}{lcccc}
			 & \multicolumn{1}{p{1.8cm}}{$\ $}	& \multicolumn{1}{p{1.8cm}}{$\ $} 
			 & \multicolumn{1}{p{1.8cm}}{$\ $} & \multicolumn{1}{p{1.8cm}}{$\ $} \\	
			 \hline \hline
			\multicolumn{1}{r}{\textsc{Age in Years:}}
			& \textsc{3 Years} & \textsc{4 Years} 
			& \textsc{5 Years} & \textsc{6 Years} \\ \cline{2-5}
			& (1) 	& (2) & (3) & (4) \\ 
			\hline 
			\multicolumn{5}{l}{\textit{Panel A:  without covariates}} \\ 
			\input out/T1-firststage1.tex
			\hline 
			\multicolumn{5}{l}{\textit{Panel B:  covariate-adjusted}} \\
			\input out/T1-firststage2.tex
			\hline 
			\input out/T1-firststageinfo.tex
			\hline 
			\multicolumn{5}{p{12.4cm}}{\footnotesize{All regressions estimated via OLS.  Standard errors (clustered at the household level) in parentheses, p-values in brackets.  Dependent variable is indicator for being enrolled in school.  Covariates included in Panel B:  school catchment area (i.e.~community) fixed effects, child age in months (fixed effects), child gender (indicator for male), child height-for-age z-score, a dummy for having an imputed value of the height-for-age z-score, an indicator equal to one if a child's mother is their primary caregiver, mother's education, an indicator for having a Luo mother, father's education, an indicator for having a Luo father, an indicator for father's presence in the household, and an indicator for missing data on the child's father, household size, the number of older siblings in the household, and a household wealth index.}} 
		\end{tabular}
	\end{small}
\end{center}
\end{table}


\begin{table}
	\caption{Distance from School and  Child Development} \label{tab:rfluo}
	\begin{center}
		\begin{small}
			\renewcommand{\arraystretch}{1.6}
			\begin{tabular}{lcccc}
				& \multicolumn{1}{p{2.1cm}}{$\ $}	& \multicolumn{1}{p{2.1cm}}{$\ $} 
				& \multicolumn{1}{p{2.1cm}}{$\ $} & \multicolumn{1}{p{2.1cm}}{$\ $} \\	[-2ex]
				\hline \hline
				\multicolumn{1}{r}{\textsc{Age in Years:}}
				& \textsc{3 Years} & \textsc{4 Years} 
				& \textsc{5 Years} & \textsc{6 Years} \\ \cline{2-5}
				& (1) 	& (2) & (3) & (4) \\ 
				\hline 
				\multicolumn{5}{l}{\textit{Panel A:  dependent variable is early childhood development index}} \\ 
				\input out/T2A-rfecd.tex
				\hline 
				\multicolumn{5}{l}{\textit{Panel B:  dependent variable is Luo receptive vocabulary index}} \\ 
				\input out/T2B-rfluo.tex
				\hline 
				\multicolumn{5}{l}{\textit{Panel C:  dependent variable is English receptive vocabulary index}} \\ 
				\input out/T2C-rfeng.tex
				\hline 
				\multicolumn{5}{l}{\textit{Panel D:  dependent variable is expressive vocabulary index}} \\ 
				\input out/T2D-rfexp.tex
				\hline 
				\multicolumn{5}{l}{\textit{Panel E:  dependent variable is fine motor skills index}} \\ 
				\input out/T2E-rfmotor.tex
				\hline 
				\input out/T2-rfinfo.tex
				\hline 
				\multicolumn{5}{p{13.4cm}}{\footnotesize{All regressions estimated via OLS.  Standard errors (clustered at the household level) in parentheses, p-values in brackets.  All specifications include controls for school catchment area (i.e.~community) fixed effects, child age in months (fixed effects), child gender (indicator for male), child height-for-age z-score, an indicator equal to one if a child's mother is their primary caregiver, mother's education, an indicator for having a Luo mother, father's education, an indicator for having a Luo father, an indicator for father's presence in the household, and an indicator for missing data on the child's father, household size, the number of older siblings in the household, and a household wealth index.}} 
			\end{tabular}
		\end{small}
	\end{center}
\end{table}



\begin{table}
	\caption{2SLS Estimates of the Impact of Preprimary Enrollment on Child Outcomes} \label{tab:iv}
\begin{center}
	\begin{small}
		\renewcommand{\arraystretch}{1.6}
		\begin{tabular}{lccccc}
			& \multicolumn{1}{p{2cm}}{$\ $} 
			& \multicolumn{1}{p{2cm}}{$\ $} & \multicolumn{1}{p{2cm}}{$\ $} 
			& \multicolumn{1}{p{2cm}}{$\ $} & \multicolumn{1}{p{2cm}}{$\ $}  \\ 
			\hline \hline
			& \multicolumn{3}{c}{\textsc{Vocabulary}} & & \\  \cline{2-4}
			& \textsc{Luo} & \textsc{English} 
			& \textsc{Expressive} & \textsc{Fine Motor} & \textsc{ECD Index}  \\  \cline{2-6}
			& (1) 	& (2) & (3)  & (4) & (5) \\ 
			\hline 
			\multicolumn{6}{l}{\textit{Panel A:  without covariates}} \\ 
			\input out/T3-iv1.tex
			\hline 	
			\multicolumn{6}{l}{\textit{Panel B:  covariate-adjusted}} \\ 
			\input out/T3-iv2.tex
			\hline 
			\input out/T3-ivinfo.tex
			\hline 
			\multicolumn{6}{p{15.8cm}}{\footnotesize{All specifications estimated via 2-stage least squares (2SLS).  First-stage F-statistics:  18.28 (Panel A) and 11.02 (Panel B).  Standard errors (clustered at the household level) in parentheses, p-values in brackets.  Covariates included in Panel B:  school catchment area (i.e.~community) fixed effects, child age in months (fixed effects), child gender (indicator for male), child height-for-age z-score, a dummy for having an imputed value of the height-for-age z-score, an indicator equal to one if a child's mother is their primary caregiver, mother's education, an indicator for having a Luo mother, father's education, an indicator for having a Luo father, an indicator for father's presence in the household, and an indicator for missing data on the child's father, household size, the number of older siblings in the household, and a household wealth index.}} 
		\end{tabular}
	\end{small}
\end{center}
\end{table}





\begin{table}
	\caption{The Impact of Distance on Likelihood of Being at Appropriate Grade-for-Age} \label{tab:gradeforage}
	\begin{center}
		\begin{small}
			\renewcommand{\arraystretch}{1.6}
			\begin{tabular}{lcccc}
				& \multicolumn{1}{p{1.8cm}}{$\ $}	& \multicolumn{1}{p{1.8cm}}{$\ $} 
				& \multicolumn{1}{p{1.8cm}}{$\ $} & \multicolumn{1}{p{1.8cm}}{$\ $} \\	
				\hline \hline
				\multicolumn{1}{r}{\textsc{Age in Years:}}
				& \textsc{3 Years} & \textsc{4 Years} 
				& \textsc{5 Years} & \textsc{6 Years} \\ \cline{2-5}
				& (1) 	& (2) & (3) & (4) \\ 
				\hline 
				\multicolumn{5}{l}{\textit{Panel A:  without covariates}} \\ 
				\input out/T4-grade4age1.tex
				\hline 
				\multicolumn{5}{l}{\textit{Panel B:  covariate-adjusted}} \\
				\input out/T4-grade4age2.tex
				\hline 
				\input out/T4-grade4ageinfo.tex
				\hline 
				\multicolumn{5}{p{15.4cm}}{\footnotesize{All regressions estimated via OLS.  Standard errors (clustered at the household level) in parentheses, p-values in brackets.  Dependent variable is indicator for being at or above appropriate grade-for age (i.e.~in first year of preprimary at age three, in second year of preprimary at age four, in third year of preprimary at age five, and in standard one at age six).  Covariates included in Panel B:  school catchment area (i.e.~community) fixed effects, child age in months (fixed effects), child gender (indicator for male), child height-for-age z-score, a dummy for having an imputed value of the height-for-age z-score, an indicator equal to one if a child's mother is their primary caregiver, mother's education, an indicator for having a Luo mother, father's education, an indicator for having a Luo father, an indicator for father's presence in the household, and an indicator for missing data on the child's father, household size, the number of older siblings in the household, and a household wealth index.}} 
			\end{tabular}
		\end{small}
	\end{center}
\end{table}


\clearpage

\appendix
\renewcommand{\thepage}{\thesection\arabic{page}}
\setcounter{page}{1}
\section{Online Appendix:  not for print publication}
\renewcommand{\thefigure}{\thesection\arabic{figure}}
\renewcommand{\thetable}{\thesection\arabic{table}}
\setcounter{table}{0}
\setcounter{figure}{0}


\begin{figure}[h]
	\caption{School Enrollment by Age} \label{fig:enroll-bars}
	\begin{center}
		\includegraphics[width=0.56\textwidth]{out/AF1-enroll-by-age-bar.pdf}
	\end{center} 
	
	\noindent
	\footnotesize{Data from 2,869 children aged 24 to 83 months in 2,013 households in 75 rural villages in western Kenya.  Bars indicate the proportion of children in each 6 month age bin who were enrolled in school, based on parent reports.}
\end{figure}

\begin{figure}
	\caption{Probability of Enrollment in Local School Among Three and Four Year Olds} \label{fig:enroll-map}
	\begin{center}
		\includegraphics[width=\textwidth]{out/AF2-id-bullseye.pdf}
	\end{center}
	
	\noindent
	\footnotesize{Data from 1,188 households with a child aged three (left panel) or four (right panel).  Figure plots the GPS location of each household, relative to the GPS location of the primary school.  Households in the first tercile of distance are less than 359.2 meters from their local school; those in the second tercile are between 359.3 and 532.5 meters from their local school, and those in the furthest tercile are between 532.7 and 750 meters from their local school.}
\end{figure}


\begin{table}
	\renewcommand{\arraystretch}{1.6}
	\caption{Summary Statistics on Three Year Olds} \label{tab:summstats}
	\begin{center}
		\begin{small}
			\begin{tabular}{lcccccc}
				\hline \hline
				& \textsc{Mean} & \textsc{S.D.} & \textsc{Median} & \textsc{Min.} & \textsc{Max.}  & \textsc{N} \\
				\hline
				\input out/A1-summstats3yrs.tex
				\hline
				\multicolumn{7}{p{13.2cm}}{\footnotesize{Data on 634 children aged 36 to 47 months.  Children are from 622 unique households (12 households include two three-year-old children). \textsc{Asset index} is the sum of indicators for having a cement floor, iron roof, latrine, or connection to the electricity grid, and indicators for owning a motorized vehicle, a bicycle, a television, a mobile phone, a computer, or a radio.}} 
			\end{tabular}
		\end{small}
	\end{center}
\end{table}


\begin{table}[h]
	\renewcommand{\arraystretch}{1.6}
	\caption{Summary Statistics on All Children Age Three to Six} \label{tab:summstats-all}
	\begin{center}
		\begin{small}
			\begin{tabular}{lcccccc}
				\hline \hline
				 & \textsc{Mean} & \textsc{S.D.} & \textsc{Median} & \textsc{Min.} & \textsc{Max.}  & \textsc{N} \\
				\hline
				\input out/A2-summstatsALL.tex
				\hline
				\multicolumn{7}{p{13.6cm}}{\footnotesize{Data on 2,503 children aged 36 to 83 months.  Children are from 1,994 unique households. \textsc{Asset index} is the sum of indicators for having a cement floor, iron roof, latrine, or connection to the electricity grid, and indicators for owning a motorized vehicle, a bicycle, a television, a mobile phone, a computer, or a radio.}} 
			\end{tabular}
		\end{small}
	\end{center}
\end{table}


\begin{table}
	\caption{Distance from School and Child Development (Without Controls)} \label{tab:rfluo-nc}
	\begin{center}
		\begin{small}
			\renewcommand{\arraystretch}{1.6}
			\begin{tabular}{lcccc}
				& \multicolumn{1}{p{2.1cm}}{$\ $}	& \multicolumn{1}{p{2.1cm}}{$\ $} 
				& \multicolumn{1}{p{2.1cm}}{$\ $} & \multicolumn{1}{p{2.1cm}}{$\ $} \\	[-2ex]
				\hline \hline
				\multicolumn{1}{r}{\textsc{Age in Years:}}
				& \textsc{3 Years} & \textsc{4 Years} 
				& \textsc{5 Years} & \textsc{6 Years} \\ \cline{2-5}
				& (1) 	& (2) & (3) & (4) \\ 
				\hline 
				\multicolumn{5}{l}{\textit{Panel A:  dependent variable is early childhood development index}} \\ 
				\input out/A3A-rfecd.tex
				\hline 
				\multicolumn{5}{l}{\textit{Panel B:  dependent variable is Luo receptive vocabulary index}} \\ 
				\input out/A3B-rfluo.tex
				\hline 
				\multicolumn{5}{l}{\textit{Panel C:  dependent variable is English receptive vocabulary index}} \\ 
				\input out/A3C-rfeng.tex
				\hline 
				\multicolumn{5}{l}{\textit{Panel D:  dependent variable is expressive vocabulary index}} \\ 
				\input out/A3D-rfexp.tex
				\hline 
				\multicolumn{5}{l}{\textit{Panel E:  dependent variable is fine motor skills index}} \\ 
				\input out/A3E-rfmotor.tex
				\hline 
				\input out/A3-rfinfo.tex
				\hline 
				\multicolumn{5}{p{13.4cm}}{\footnotesize{All regressions estimated via OLS.  Standard errors (clustered at the household level) in parentheses, p-values in brackets.  All specifications include controls for school catchment area (i.e.~community) fixed effects, child age in months (fixed effects), child gender (indicator for male), child height-for-age z-score, an indicator equal to one if a child's mother is their primary caregiver, mother's education, an indicator for having a Luo mother, father's education, an indicator for having a Luo father, an indicator for father's presence in the household, and an indicator for missing data on the child's father, household size, the number of older siblings in the household, and a household wealth index.}} 
			\end{tabular}
		\end{small}
	\end{center}
\end{table}


\begin{table}
	\renewcommand{\arraystretch}{1.6}
	\caption{Does Distance Predict Observable Characteristics of Children and Households?} \label{tab:exclusion}
	\begin{center}
		\begin{small}
			\begin{tabular}{lccc}
				& \multicolumn{1}{p{1.4cm}}{$\ $} 
				& \multicolumn{1}{p{1.4cm}}{$\ $} 
				& \multicolumn{1}{p{1.4cm}}{$\ $} \\ [-2ex]
				\hline \hline
				& \textsc{Coefficient} & \textsc{S.E.} & \textsc{p-value} \\
				\hline 
				\input out/A4-ivcheck.tex
				\hline 
				\multicolumn{4}{p{11.4cm}}{\footnotesize{Coefficients from OLS regressions of outcome variables on distance from the school (in km).  Data on 634 children aged 36 to 47 months.  Children are from 622 unique households (12 households include two three-year-old children).  Standard errors clustered at the household level.  \textsc{Asset index} is the sum of indicators for having a cement floor, iron roof, latrine, or connection to the electricity grid, and indicators for owning a motorized vehicle, a bicycle, a television, a mobile phone, a computer, or a radio.}} 
			\end{tabular}
		\end{small}
	\end{center}
\end{table}


\begin{table}
	\caption{Complier Characteristics:  OLS Regressions of School Enrollment on Distance} \label{tab:compliers}
	\begin{center}
		\begin{small}
			\renewcommand{\arraystretch}{1.6}
			\begin{tabular}{lcccc}
				& \multicolumn{1}{p{2.4cm}}{$\ $} 
				& \multicolumn{1}{p{2.4cm}}{$\ $} & \multicolumn{1}{p{2.4cm}}{$\ $} 
				& \multicolumn{1}{p{2.4cm}}{$\ $} \\ 
				\hline \hline
				& & 	\multicolumn{3}{c}{\textsc{Indicator for Below Median:}} \\ \cline{3-5}
				\multicolumn{1}{r}{\textit{Characteristic:}} & \textsc{Male} & \textsc{HAZ} 
				& \textsc{Assets} & \textsc{Mother's Ed.} \\  \cline{2-5}
				& (1) 	& (2) & (3)  & (4) \\ 
				\hline 
				\input out/A5-compliers.tex
				\hline 
				\hline 
				\multicolumn{5}{p{15cm}}{\footnotesize{All specifications estimated via OLS.  Standard errors (clustered at the household level) in parentheses, p-values in brackets.  The outcome variable in all specifications is an indicator for school enrollment.  The sample is restricted to three year olds.  All specifications include the following covariates:  school catchment area (i.e.~community) fixed effects,  child age in months (fixed effects), child gender (indicator for male), child height-for-age z-score, a dummy for having an imputed value of the height-for-age z-score, an indicator equal to one if a child's mother is their primary caregiver, mother's education, an indicator for having a Luo mother, father's education, an indicator for having a Luo father, an indicator for father's presence in the household, and an indicator for missing data on the child's father, household size, the number of older siblings in the household, and a household wealth index.}} 
			\end{tabular}
		\end{small}
	\end{center}
\end{table}



\begin{table}
	\caption{Cross-Sectional Relationship Between Preprimary Enrollment and Outcomes} \label{tab:biasedols}
\begin{center}
	\begin{small}
		\renewcommand{\arraystretch}{1.6}
		\begin{tabular}{lccccc}
			& \multicolumn{1}{p{2cm}}{$\ $} 
			& \multicolumn{1}{p{2cm}}{$\ $} & \multicolumn{1}{p{2cm}}{$\ $} 
			& \multicolumn{1}{p{2cm}}{$\ $} & \multicolumn{1}{p{2cm}}{$\ $}  \\ 
			\hline \hline
			& \multicolumn{3}{c}{\textsc{Vocabulary}} & & \\  \cline{2-4}
			& \textsc{Luo} & \textsc{English} 
			& \textsc{Expressive} & \textsc{Fine Motor} & \textsc{ECD Index}  \\  \cline{2-6}
			& (1) 	& (2) & (3)  & (4) & (5) \\ 
			\hline 
			\multicolumn{6}{l}{\textit{Panel A:  without covariates}} \\ 
			\input out/A6A-biasedols1.tex
			\input out/A6-biasedolsinfo_r2noc.tex
			\hline 	
			\multicolumn{6}{l}{\textit{Panel B:  covariate-adjusted}} \\ 
			\input out/A6B-biasedols2.tex
			\input out/A6-biasedolsinfo_r2c.tex
			\hline 
			\input out/A6-biasedolsinfo_obs.tex
			\hline 
			\multicolumn{6}{p{15.8cm}}{\footnotesize{All cross-sectional specifications estimated via OLS. A separate regression for each outcome is presented in each column of each panel. Child outcomes are listed at the top of each column. Standard errors (clustered at the household level) in parentheses, p-values in brackets.  As in other estimates, the sample includes 634 children aged 36 to 47 months.  Covariates included in Panel B:  school catchment area (i.e.~community) fixed effects, child age in months (fixed effects), child gender (indicator for male), child height-for-age z-score, a dummy for having an imputed value of the height-for-age z-score, an indicator equal to one if a child's mother is their primary caregiver, mother's education, an indicator for having a Luo mother, father's education, an indicator for having a Luo father, an indicator for father's presence in the household, and an indicator for missing data on the child's father, household size, the number of older siblings in the household, and a household wealth index.}} 
		\end{tabular}
	\end{small}
\end{center}
\end{table}

\end{document}